Figure 6
From: Using deep neural networks as a guide for modeling human planning
Representative residuals between the baseline model and the data (first column) and the baseline model and the neural network (second column). For each board position, we report the KL divergence between the output distributions of the model and the network, as well as the negative log-likelihood of the human move for the model and the network. The user is playing black while the computer opponent is playing white. Additionally, the red shading indicates the probability distribution of the network’s move prediction, the open circle indicates the user’s selected move, and the dashed circle indicates the baseline model’s predicted move.
